We are customer-centric and provide customized or one-stop full-stack solutions to empower all industries.
The oil industry should be one of the most widely used fields for HPC high-performance computing, because oil exploration requires intensive calculation and simulation of massive data, and the calculation results must be converted into intuitive visual images. These tasks must rely on high-performance computers to achieve the best exploration benefits. In oil exploration projects, geophysical methods are needed to obtain data, including electrical, magnetic, gravity, radioactive and seismic wave methods. Seismic wave method is the most commonly used basic method in oil exploration, because seismic waves can penetrate rock formations several thousand meters thick. Provide information and data on possible oil burials, and then use the powerful data processing capabilities of HPC parallel computing to accurately calculate and analyze the acquired information and data to provide reference for oil and gas drilling positioning.
In order to achieve stronger computing and analysis capabilities for massive data, high requirements are placed on the design of communication network architecture, parallel computing architecture, management structure, and storage architecture:
1) Satisfy the continuous bandwidth requirements for temporary computing data processing speed due to fast computing
2) Changes in computing architecture and computing power. Traditional host-based computing deployment architecture has shortcomings
3) Comprehensively consider the characteristics of high-performance computing applications in the oil industry, as well as the oil companies’ requirements for efficiency and cost control of exploration business computing tasks.
4) High availability requirements for large-scale systems in computing nodes, networks, storage, and software
5) Requirements for centralized maintenance, simplicity, convenience and reliability management of large-scale system equipment
According to the actual business processing needs of the project, this solution adopts the PowerScale9000 all-in-one machine with login nodes, management nodes and high-speed communication network centralized deployment solution. Cluster processing is performed between the service computing nodes to achieve balanced load of computing business and smooth migration; and link redundancy mechanism is used to ensure the stability and security of communication data transmission; overall meeting the needs of the project.
In this solution, PowerLeader’s high-performance customized Powerscale 9000 supercomputer (HPC) system is selected as the data processing center in the field of oil exploration, playing a pivotal role.
High density: Ultra-high density design, can support up to 80 nodes, saving space
Easy to expand: adopt platform design concept and fully modular design
Low power consumption: centralized power supply and shared heat dissipation design
Easy to manage: integrated and fast delivery design
Easy maintenance: Centralized management design is adopted to simplify operation and maintenance management and improve operation and maintenance efficiency
Cluster deployment: Based on the parallel computing architecture, all settlement nodes and management nodes are connected through a high-speed communication network, and cluster software is deployed to achieve load balancing of tasks. When the computing nodes process tasks unevenly, tasks are automatically and equally distributed to achieve equal computing effects, thus achieving fast and efficient parallel computing. Deploy centralized storage methods, with each node mapping the same storage to achieve data sharing, convenient and flexible calling, storage and writing of data information, and deploy centralized management nodes to centrally manage and monitor the status and operation of each computing node.
Network deployment: 40GB and 80GB INFINIBAND switches are used to interconnect computing nodes, management nodes, storage nodes, and login nodes, achieving high-bandwidth, high-speed, and low-latency data communication, and solving the bottleneck of IO communication. Link redundancy mechanism is used to prevent single-channel failures, single-channel blockages that lead to decreased computing performance, interruptions, and low speed.
Node deployment
Deploy four nodes in the cluster according to the application and function
Management Node
Storage Node
calculate node
Login Node
Management Node is the control node of various management measures of the cluster system, the control point of the management network, and monitors the operation status of each node and network in the cluster. The cluster management software runs on this node. Storage Node Storage node is the data storage and data server of the cluster system. It needs to store a large amount of data, deploy parallel file systems and multiple IO servers. Computing Node is the computing core of the entire cluster system. Its function is to perform calculations. It uses multi-channel multi-core servers or blades and multi-node servers to achieve high performance, high density, low energy consumption, and parallel fast calculations. Login node is responsible for user information registration, login information verification, system use, and operation security mechanism.
The entire system can be centrally managed through the business network, making maintenance easy for customers.
CNOOC systems have always used foreign server brands, for which they have to pay high software support fees every year. The use of domestic HPC equipment can not only fully meet the work needs of oil exploration, but also has low costs and good user experience.
Cluster software is used to achieve unified cluster deployment, automatically balance loads, and migrate data to ensure reliable computing operations. High-bandwidth, high-speed INFINIBAND switches are used to completely resolve low-speed IO bottlenecks.
The communication link adopts a redundant design, and redundant links are used to connect the computing nodes, management nodes and storage nodes to ensure safe and smooth data reading and writing.
